Phase and group velocity measurements from physically modeled transmission gathers

Similar documents
Estimation of stiffness coefficients of an orthorhombic physical model from group velocity measurements

AVAZ inversion for fracture orientation and intensity: a physical modeling study

Robert W. Vestrum and R. James Brown

AVAZ inversion for fracture orientation and intensity: a physical modeling study

Observation of shear-wave splitting from microseismicity induced by hydraulic fracturing: A non-vti story

Finite difference elastic modeling of the topography and the weathering layer

GROUP- AND PHASE-VELOCITY INVERSIONS GENERAL ANISOTROPIC STIFFNESS TENSOR FOR THE. Robert W. Vestrum THE UNIVERSITY OF CALGARY

Numerical Modeling for Different Types of Fractures

Salt anisotropy: Ultrasonic lab experiments and traveltime ramifications Jingjing Zong *, Robert Stewart and Nikolay Dyaur University of Houston

Acquisition and preliminary analysis of the Castle Mountain shallow VSP dataset

Radiation pattern in homogeneous and transversely isotropic attenuating media

An acoustic wave equation for orthorhombic anisotropy

6298 Stress induced azimuthally anisotropic reservoir - AVO modeling

Azimuthal Velocity Analysis of 3D Seismic for Fractures: Altoment-Bluebell Field

PEAT SEISMOLOGY Lecture 9: Anisotropy, attenuation and anelasticity

The signature of attenuation and anisotropy on AVO and inversion sensitivities

Azimuthal AVO and Curvature. Jesse Kolb* David Cho Kris Innanen

Effect of fracture scale length and aperture on seismic wave propagation: An experimental study

Relationships between the velocities and the elastic constants of an anisotropic solid possessing orthorhombic symmetry

Automatic time picking and velocity determination on full waveform sonic well logs

An investigation of the free surface effect

SEISMIC WAVE PROPAGATION IN FRACTURED CARBONATE ROCK

Application of Seismic Reflection Surveys to Detect Massive Sulphide Deposits in Sediments-Hosted Environment

Drift time estimation by dynamic time warping

AVAZ inversion for fracture orientation and intensity: A physical modeling study. Faranak Mahmoudian Gary Margrave

Snell s law in transversely isotropic media using linearized group velocities and related quantities

Borehole Geophysics. Acoustic logging measurements

Phase-shift modelling for HTI media

Chapter 1. Introduction EARTH MODEL BUILDING

Receiver. Johana Brokešová Charles University in Prague

Reflection Seismic Method

Elastic wavefield separation for VTI media

Static Corrections for Seismic Reflection Surveys

Source-geophone azimuth from 3-C seismic polarization

Physical modeling and analysis of P-wave attenuation anisotropy in transversely isotropic media

Seismic anisotropy and salt detection: A physical modelling study. Zandong Sun, R.James Brown, and Don C. Lawton

Introduction to Seismology Spring 2008

Th Guided Waves - Inversion and Attenuation

Seismic inversion for the parameters of two orthogonal fracture sets in a VTI background medium

Inversion of normal moveout for monoclinic media

A Case Study on Simulation of Seismic Reflections for 4C Ocean Bottom Seismometer Data in Anisotropic Media Using Gas Hydrate Model

Correspondence between the low- and high-frequency limits for anisotropic parameters in a layered medium

The Deconvolution of Multicomponent Trace Vectors

A Petroleum Geologist's Guide to Seismic Reflection

Seismic Waves in Complex 3 D Structures, 26 (2016), (ISSN , online at

NMO ellipse for a stratified medium with laterally varying velocity

Reflection moveout and parameter estimation for horizontal transverse isotropy

Detecting fractures using time-lapse 3C-3D seismic data

Borehole Acoustics and Logging Consortium. Annual Report

P137 Our Experiences of 3D Synthetic Seismic Modeling with Tip-wave Superposition Method and Effective Coefficients

An efficient wave extrapolation method for tilted orthorhombic media using effective ellipsoidal models

F-K Characteristics of the Seismic Response to a Set of Parallel Discrete Fractures

Interpretation of baseline surface seismic data at the Violet Grove CO 2 injection site, Alberta

Absolute strain determination from a calibrated seismic field experiment

Estimation of fractured weaknesses and attenuation factors from azimuthal seismic data

Effects of Fracture Parameters in an Anisotropy Model on P-Wave Azimuthal Amplitude Responses

Velocity and VTI anisotropy scanning in multicomponent seismic data

An empirical method for estimation of anisotropic parameters in clastic rocks

Conversion coefficients at a liquid/solid interface

Seismic applications in coalbed methane exploration and development

Acoustic Anisotropy Measurements and Interpretation in Deviated Wells

Towards Modelling Elastic and Viscoelastic Seismic Wave Propagation in Boreholes

Body-wave radiation patterns and AVO in transversely isotropic media

A collision theory of seismic waves applied to elastic VSP data

The effect of anticlines on seismic fracture characterization and inversion based on a 3D numerical study

A new S-wave seismic source

AVAZ and VVAZ practical analysis to estimate anisotropic properties

Far-field radiation from seismic sources in 2D attenuative anisotropic media

Time lapse view of the Blackfoot AVO anomaly

P-wave reflection coefficients for transversely isotropic models with vertical and horizontal axis of symmetry

Ground-penetrating radar (GPR) and shallow seismic surveys at Calgary International Airport, Alberta

Extensional and Flexural Waves in a Thin-Walled Graphite/Epoxy Tube * William H. Prosser NASA Langley Research Center Hampton, VA 23665

3D Converted Wave Data Processing A case history

6th NDT in Progress Lamb waves in an anisotropic plate of a single crystal silicon wafer

Reflection and Transmission coefficient for VTI media

The exact eikonals (Hamiltonians) of the coupled quasi-compressional ( qp ) and quasi-shear ( qs

7.2.1 Seismic waves. Waves in a mass- spring system

Kirchhoff prestack depth migration in simple models with differently rotated elasticity tensor: orthorhombic and triclinic anisotropy

Analysis of multicomponent walkaway vertical seismic profile data

Mapping the conversion point in vertical transversely isotropic (VTI) media

Elastic wave-equation migration for laterally varying isotropic and HTI media. Richard A. Bale and Gary F. Margrave

SPECTRAL FINITE ELEMENT METHOD

Improvement of stacking image by anisotropic velocity analysis using P-wave seismic data

An Improvement of Velocity Variation with Offset (VVO) Method in Estimating Anisotropic Parameters

P235 Modelling Anisotropy for Improved Velocities, Synthetics and Well Ties

APPLICATION-DIRECTED MODELING OF RADIATION AND PROPAGATION OF ELASTIC WAVES IN ANISOTROPIC MEDIA: GPSS AND OPOSSM

P306 Seismic Velocity Anisotropy in the Illizi Basin of Eastern Algeria

Seismic tests at Southern Ute Nation coal fire site

Lamb Waves in Plate Girder Geometries

Microseismic data illuminate fractures in the Montney

INTRODUCTION TO APPLIED GEOPHYSICS

We Challenges in shale-reservoir characterization by means of AVA and AVAZ

Elements of 3D Seismology Second Edition

Nonhyperbolic Reflection Moveout for Orthorhombic Media

Orthorhombic anisotropy: A physical model study. Scott P. Cheadle, R. James Brown and Don C. Lawton

Laboratory experiments and numerical simulation on Bitumen Saturated Carbonates: A Rock Physics Study for 4D Seismology

SHEAR-WAVE REFLECTION MOVEOUT FOR AZIMUTHALLY ANISOTROPIC MEDIA

Elastic wave-mode separation for VTI media

Velocity Dispersion and Attenuation in Vibroseis Data

Case History. 3-D AVO analysis and modeling applied to fracture detection in coalbed methane reservoirs. Antonio C. B. Ramos and Thomas L.

Transcription:

Phase and group velocity measurements Phase and group velocity measurements from physically modeled transmission gathers Faranak Mahmoudian, Gary Margrave, and Joe Wong ABSTRACT Physical model data have been used for many years to simulate exploration targets, as in the example of a fractured medium. Yet, physical modeling is challenging for at least two reasons, (1) the initial characterization of the medium is difficult, and (2) the large highly-directional transducers used as sources and receivers cause distortions. The initial characterization of a laboratory physical model is done by determination of elastic stiffness coefficients from the phase or group velocity measurements along various directions. We present a review on how to measure phase and group velocities from physical model transmission gathers acquired using piezoelectric transducers with different sizes. Group velocity measurements are found to be straightforward, reasonably accurate, and independent of the size of the transducers used. In contrast, the accuracy of phase velocities derived from the (τ, p) transform analysis was found to be very sensitive to small differences in picked arrival times and to transducer size. Compared to the phase-velocity procedure, the technique involving group velocities is much less prone to error due to time-picking uncertainties, and therefore is more suitable for analyzing physical model seismic data. INTRODUCTION In physical seismic modeling, an alternative to numerical modeling, the seismic data are acquired over small, laboratory sized geological models. Physical modeling has been used to evaluate the accuracy of mathematical models of wave propagation, to test seismic data processing algorithms, and to provide insights into the interpretation of 3C seismic data acquired over complex media (Ebrom and McDonald, 1994). Further, it has acquired relevance as a method for studying anisotropic media, because waves in laboratory scale models obey the same physical laws as waves propagating in the anisotropic earth. In physical modeling of anisotropic media, fractured media in particular, the first matter to be addressed is the construction of a model that is reasonably representative of a geological feature, with known anisotropic elastic properties. Knowing the 6 6 stiffness coefficient matrix and density of the laboratory model determines all its elastic properties. The stiffness coefficients can be estimated from phase or group velocities of the three quasi-body wave types; a quasi-compressional (qp) and two quasi-shear (qs) modes. There are well-known exact explicit relations between the anisotropic stiffness coefficients, in particular for orthorhombic symmetry, and the quasi-body wave s phase velocities, obtained from the solution of the Christoffel equation (Musgrave, 1970; Tsvankin, 1997). Consequently, the anisotropic stiffness coefficients are most often determined by measuring phase velocities (e.g., McSkimin, 1967; Vestrum et al., 1999; Mah and Schmitt, 2001). The phase velocity approach leaves no doubt as to which of the stiffness coefficients can be recovered, and the uncertainties in the value of stiffness coefficients can usually be related in a simple way to the experimental errors in the phase velocity measurements. For large physical models, however, measurements of phase velocities are problematic, require special ex- CREWES Research Report Volume 25 (2013) 1

Mahmoudian et al. perimental setups if the (τ, p) transform method is employed, dependent upon the size of transducers used, and prone to large errors. A straightforward alternative to phase velocity methods is employing the group velocity which controls the traveltime and is easy to measure and less prone to errors. For group velocity, exact explicit relationships to the stiffness coefficients are not well-known, which can be manipulated to obtain the stiffness coefficients directly. The existing methods for recovering stiffness coefficients from group velocities employ indirect procedures to obtain the stiffness coefficients from group velocity measurements, for more explanation see Mahmoudian et al. (2013). In this report, using numerical simulation, the methods for measuring both phase and group velocities from physical modeling transmission gathers are explained and compared. The group velocity measurements are found to be straightforward, reasonably accurate, and independent of the size of the transducers used. In contrast, the phase velocity measurements require specialized acquisition geometries, are prone to large measurement errors, and are quite sensitive to the size of the transducers used. Over an anisotropic model, we measured the phase and group velocities along various directions from physical model transmission gathers. The group velocity measurements were then used to estimate the density-normalized stiffness coefficients of the phenolic model which is presented elsewhere Mahmoudian et al. (2013). The group velocity measurement from physical model data is straightforward, accurate, independent of the size of the transducers used (hence practical), and is particularly useful for initial characterization of an orthorhombic physical layer, required in advance of using the layer in any physical modeling seismic survey. BACKGROUND Seismic anisotropy is the dependence of elastic wave velocities through a medium on the direction in which the elastic wave is traveling. The phase velocity is the wavefront s propagation velocity in the direction normal to the wavefront, and the group (ray) velocity is the energy propagation velocity. Consider the reference Cartesian coordinate system, (x 1, x 2, x 3 ), associated with the three orthorhombic symmetry planes. Along the coordinate principal directions (i.e. x 1 -, x 2 -, and x 3 -axes), phase and group velocities are equal. In this reference coordinate system, the nine independent orthorhombic density-normalized stiffness coefficients are the six diagonal A ii plus three off-diagonal terms (A 23, A 13, A 12 ). The quasi-body wave velocities along the coordinate principal axes determine diagonal stiffness coefficients (Table 1), the A ii (i = 1 : 3) from the three quasi-p (qp) velocities, and the A ii (i = 4 : 6) from the three quasi-s (qs) velocities. The off-diagonal stiffness coefficients, however, are not individually related to the phase or group velocity along some particular directions. Determination of the off-diagonal density-normalized stiffness coefficients from group velocity measurements are explained in Mahmoudian et al. (2013). VELOCITY MEASUREMENTS For rock samples, seismic velocities are often measured using relatively large ultrasonic transducers in a transmission geometry (Bouzidi and Schmitt, 2009). Using two flat-faced transducers attached to the model in various propagation directions, body-wave velocities can be measured by picking arrival times. According to Dellinger and Vernik (1994) and 2 CREWES Research Report Volume 25 (2013)

Phase and group velocity measurements Propagation Polarization x 1 x 2 x 3 x 1 V 11 = A 11 V 12 = A 66 V 13 = A 55 x 2 V 21 = A 66 V 22 = A 22 V 23 = A 44 x 3 V 31 = A 55 V 32 = A 44 V 33 = A 33 Table 1. Body waves velocities along the principal axes. Here V ij (i, j = 1, 2, 3) is the body wave velocity which propagates along the x j -axis and is polarized along the x i -axis. For example V 11 is the qp velocity propagating along the x 1 -axis, and V 23 (= V 32 ) is the qs velocity propagating along the x 3 -axis and polarized along the x 2 -axis. FIG. 1. A transmission gather using large transducers, modeled by an acoustic finite-difference method, over a model with the size of 160 80mm 2. The black parallel vectors show the travel path of the plane-wave portion of the wavefront. The transducer s size is chosen to be half of the layer thickness to exaggerate the plane-wave generation. Vestrum (1994) such measurements estimate group velocities if the source-receiver separation is large compared to the transducer dimension. If the transducers are large compared to their separation, they will approximately transmit and receive plane waves over a large spatial interval (Figure 1), thus enabling direct phase velocity measurement. For a small specimen to measure phase velocity using large transducers, the specimen is cut appropriately along various directions to make the desired contact plane for the flat-faced large transducers. The phase velocity along the direction normal to the cut planes then can be measured. Next, we describe measuring group and phase velocity on large physical models which simulate some geological features rather than on small core samples. Group velocity NUMERICAL SIMULATIONS Consider a physical model transmission gather, where the source transducer is located on one side of the model and the receiver profile is positioned on the other side. Treating the physical model as a homogeneous plane layer, for small transducers which approximate point sources and receivers, the length of the straight line connecting the centers of the transducers divided by the first arrival traveltimes yields a good estimate of the group velocity in the source-receiver center-to-center raypath direction. For large source and receiver transducers, the effective source-receiver raypath is the straight line connecting the nearest-edges of source and receiver transducers, because the energy propagates the minimum distance between the transducers. Brown et al. (1991) calculated the group velocities in various directions by dividing such effective source-receiver raypaths by the first arrival CREWES Research Report Volume 25 (2013) 3

Mahmoudian et al. Traveltime (µs) 20 30 40 50 0 50 100 150 Offset (mm) (a) Group velocity (m/s) 3700 3600 3500 3400 3300 0 10 20 30 40 50 60 70 Group angle (degrees) (b) FIG. 2. (a) Finite-difference generated transmission gather over a constant velocity layer with a velocity of 3500m/s and thickness of 70mm and source and receiver size of 13mm. This is an 2D acoustic wave propagating through a uniform velocity grid with the grid size of 0.5mm and time step of 0.05µs. (b) Measured group velocity. traveltimes of each mode. We follow their method in measuring the group velocities. Hence the effective size of transducers needs to be known. The effective size of a transducer is defined by the active portion of the piezoelectric crystal, and may be slightly different from the nominal dimension given by the manufacturer. The appropriate effective transducer size is determined from preliminary measurements in an isotropic homogeneous plexiglas layer. The justification for considering the nearest-edges distance as the raypath taken by the first arrival energy is provided below. We assume that a circular physical model transducer can be considered to be a continuous seismic source/receiver array. Considering a source transducer, assuming Huygens s principle, a transducer can be regarded as circular array of point sources where each individual element radiates the same waveform simultaneously with the others. Using acoustic finite-difference modeling over a 70mm isotropic layer with a constant velocity of 3500m/s, we generated a transmission gather utilizing linear source and receiver arrays representing a transducer with the diameter of 13mm (Figure 2a). The finite-difference program we used, utilizes a second-order nine-point approximation to the Laplacian operator. The time steps were chosen to be small enough to ensure that grid dispersion is as small as numerically possible. A minimum phase wavelet with a dominant frequency of 500kHZ was used as the initial wavelet. In the finite-difference generated transmission gather (Figure 2a), the change in wavelet shape from near to far offsets is apparent and is due to the finite sizes of source and receivers. The 13mm source/receiver is implemented using source/receiver array, with the array length of 13mm. The wavelet change is due to the array effect. Figure 2b shows the measured group velocity, obtained from first arrival times divided by the nearest-edges distance between transducers, versus the group angle with the error of ±10m/s. Using the edge-to-edge distance rather than the center-to-center distance between the source and receiver transducers, is a simple geometrical correction. Figure 3 shows the In a linear array, every point has the same weight. Software from CREWES MatLab library. 4 CREWES Research Report Volume 25 (2013)

Phase and group velocity measurements Group velocity (m/s) 3600 3500 3400 3300 3200 3100 3000 1.1mm transducers 6mm transducers 13mm transducers 2900 0 10 20 30 40 50 60 70 Group angle (degrees) FIG. 3. Group velocity estimated from transmission data, (x 1, x 3 ) plane, acquired with three transducer sizes using the edge-to-edge distance correction. group velocities of the experimental phenolic layer measured from physical model transmission gathers acquired with 1.1mm, 6mm, and 13mm transducers after the edge-to-edge correction. The measurements from these three transducer sizes are consistent to within the picking error, and independent of the size of the transducer used. Phase velocity Kebaili and Schmitt (1997) presented a method using the (τ, p) transform to measure phase velocity from two transmission shot gathers with two different shot depths. Consider two transmission shot gathers recorded with the shots at depths of d 1 and d 2, acquired by point sources and receivers (Figure 4a). In each transmission gather the first arrival event of each mode, for example P-wave, appears as a hyperbolic event (Figure 4b). Taking the (τ, p) transform, the first arrival hyperbola will be mapped to an apparent ellipse (Figure 4c). In each transmission gather, consider the line with the slope of p 0 (t = τ 1 + p 0 x and t = τ 2 + p 0 x) tangent to the first arrival hyperbola (Figure 4b). These lines designate plane waves traveling with the horizontal slowness of p 0, intercepting the time axis at τ 1 and τ 2. Summation along these two lines produces the two points, (τ 1, p 0 ) and (τ 2, p 0 ), on each (τ, p) event, respectively. Effectively, the p 0 plane wave has traveled the vertical distance of d 2 d 1 with the vertical slowness of q 0 = (τ 2 τ 1 )/(d 2 d 1 ) from the source at d 1 to the source at d 2. Therefore, for any particular horizontal slowness of p 0 in the (τ, p) domain, the two (τ 1, p 0 ) and (τ 2, p 0 ) points, picked from the transform of the first arrival hyperbolic events, will estimate the phase velocity as v p0 = (p 0 2 + q 0 2 ) ( 1/2). For a point source and receivers, we generated two finite-difference transmission gathers through an isotropic layer with the constant velocity of 3500m/s, with the shot depths at d 1 = 40mm and d 2 = 50mm (Figure 5). For each p 0, the corresponding (τ 1, p 0 ) and (τ 2, p 0 ), are picked from the ellipses in the (τ, p) transforms of the two gathers. Figure 6a shows the measured phase velocities versus phase angle. As is apparent, the Kebaili and Schmitt (1997) method successfully measures the constant phase velocity of 3500m/s with an error of ±50m/s due to picking uncertainty. Assuming a point source and a sufficient number of point receivers, the picking in the (τ, p) transform can be done consistently on Note, pure hyperbola and ellipse events are only true for isotropic case. CREWES Research Report Volume 25 (2013) 5

Mahmoudian et al. first arrival, peak, or trough within this error range. Next, we repeated the same finite-difference transmission gathers, using source and receivers with the simulated size of 13mm. The receivers were attached to the top surface (Figure 4a), and the source transducer at the side of the layer with its top edge at the depths of d 1 and d 2. Figure 6b shows the measured phase velocity comparing the results when it was picked on the first arrival, peak or trough (Figure 5b). The measured phase velocity has the error range of ±400m/s when peak or trough was picked (Figure 6b), but seems more accurate with error range of ±150m/s when first arrival was used. We believe that such large errors are related to the loss of resolution caused by the finite-size sources and receivers, which was simulated by using source/receiver arrays. Essentially, the array effect is a spatial averaging and the resulting distortion is less for first arrivals than for subsequent arrivals. For physical model data acquired over anisotropic models, this wave interference plus the presence of noise in the data reduces the reliability of phase velocity measurements. Hence, for phase velocity measurements smaller transducers, resembling point sources and receivers, as used in Mah and Schmitt (2001), are desired. This contrasts with the results in the previous section where we showed that the measurement of group velocity is less sensitive to transducer size. PHYSICAL MODEL EXPERIMENT Our physical modeling system has a scale of (1 : 10000) for both length and time. This means that, for example, 1mm in the physical model represents 10m and a frequency of 500kHz represents a frequency of 50Hz in the real earth. Having the same scale for length and time allows the velocity of the medium to remain unscaled. We used flat-faced circular piezoelectric transducers of three different for acquiring transmission data. These compressional and shear-wave (P and S) transducers convert electrical energy to mechanical energy and vice-versa, thus being capable of acting as either sources or receivers. As a receiver, the P- and S-transducers are sensitive to displacement normal and tangential to the contact face of the transducer, respectively, and represent vertical and horizontal component geophones. As a source, either P- or S-transducers generate both P- and S-waves, but generally stronger P-wave emanates from the P-transducer and stronger S-waves emanate from the S-transducer. We took special measures in coupling S-transducers to a solid surface to decrease the P-wave amplitudes produced by S-transducers. Our modeling system is equipped with a robotic positioning system which has a positioning error less than 0.1mm. There are separate arms for both the source and receiver. Trace stacking of repeated source excitations for each receiver position and the progressive re-positioning of the receiver transducer generates a seismic gather. The location of a transducer is assigned to the location of the center of its contact face. We manually position the first source and receiver locations according to a predefined coordinate system. Once the initial source-receiver offset is set, the subsequent increments in offset are computer controlled, and as a consequence are accurately known. The source pulse is highly repeatable over many hours of acquisition. More details about the laboratory equipment and set-up are described in Wong et al. (2009). 6 CREWES Research Report Volume 25 (2013)

Phase and group velocity measurements (a) (b) (c) FIG. 4. (a) Schematic view of the acquisition geometry of the two transmission gathers required for measuring the phase velocity. (b) The two traveltime versus offset curves for the two transmission gathers with different source positions. (c) (τ, p) transforms of the traveltime curves in (b). Note the determination of τ = τ 2 τ 1 at a given horizontal slowness p 0 (Mah and Schmitt, 2001). CREWES Research Report Volume 25 (2013) 7

Mahmoudian et al. (a) (b) FIG. 5. (a) Finite-difference generated transmission gather assuming point source and receivers, with the shot depths at d 2 = 50mm, over a constant velocity layer with the velocity of 3500m/s. (b) The (τ, p) transform of the gather in (a). The first breaks picks are shown by blue dots, on peaks are shown by white dots, and on trough are shown by red dots. Two of such transmission gathers, with the shot depths at d 1 = 40mm and d 2 = 50mm, were used in the estimation of phase velocities. Phase velocity (m/s) 3700 3600 3500 3400 3300 Phase velocity (m/s) 3700 3600 3500 3400 3300 peak trough first break 0 10 20 30 40 50 60 70 Phase angle (degrees) 0 10 20 30 40 50 60 70 Phase angle (degrees) (a) (b) FIG. 6. Phase velocity measured from finite-difference data gathered over an isotropic layer with the constant velocity of 3500m/s. (a) Point source and receiver data. (b) 13mm transducer data. 8 CREWES Research Report Volume 25 (2013)

Phase and group velocity measurements x 3 x 2 x 1 (a) (b) FIG. 7. (a) The simulated fractured medium used in this study. The dark lines indicate glued seams between separate phenolic blocks. (b) A slab of phenolic material with the dashed lines displaying the linen planes. Simulated fractured layer We constructed a simulated fractured medium from phenolic LE-grade material, at the University of Calgary, CREWES Project. Phenolic LE is composed of laminated sheets of linen fabric bonded together with phenolic resin, and has mass density of 1390 kg/m 3. To construct our simulated fractured layer, the original board of phenolic material with horizontally-laid linen fabric was cut into slabs along planes orthogonal to the plane of linen layers. These were rotated 90 and bonded together under a uniform high pressure with epoxy. This constructed layer simulates a VTI medium with vertical parallel fractures or an orthorhombic media. It has an approximate area of 57 57cm 2 and a thickness of 7cm (Figure 7). Group velocity measurements EXPERIMENTAL PHENOLIC LAYER We acquired 3C transmission seismic data over the physical phenolic model described previously, which is intended to simulate a medium with parallel vertical fractures. The vertical, radial, and transverse component data were all acquired, utilizing P-transducers, radially polarized S-transducers, and transversely polarized S-transducers, as source and receivers, respectively. For each component, the source and receiver transducers always had identical polarizations. The P- and S-transducers are Panametric V 103 and V 153 with diameters of 13 mm and a scaled nominal central frequency of 32Hz for P-wave. For the acquisition, the reference Cartesian coordinate system was chosen to be the same as the orthorhombic symmetry system. As the symmetry of phenolic materials is relatively well controlled, the Cartesian axes were aligned with the symmetry planes of the phenolic layer. The transmission receiver lines were positioned along 0, 90, 45, and 135 azimuths on the model s top surface (Figure 8a-d), with the source located at the bottom; and 0, 90 azimuth receiver lines on the top surface were used with the source also on the top of the model, offset some distance from the receiver line (Figure 8e-f). The group velocities along different directions in the (x 1, x 3 ), (x 2, x 3 ), azimuth 45, and azimuth 135 planes were estimated from the transmission profiles in Figure 8a-d. The group velocities along different directions in the (x 1, x 2 ) plane were estimated from the profiles in Figure 8e-f. Figure 10 shows the 3C data from one of our transmission shot gathers. The shearwave splitting phenomenon is observed between the radial and transverse components (Figure 10c-d), with the zero-offset arrivals of the qs-waves at different arrival times (approximately 0.45ms and 0.41ms, respectively). The qp group velocities are determined CREWES Research Report Volume 25 (2013) 9

Mahmoudian et al. x 3 * x 1 (a) x 2 x 3 * x 1 (b) x 2 x 3 * x 1 (c) x 2 x 3 * x 1 (d) x 2 x 3 * x 2 x 3 * x 2 x 1 x 1 (e) (f) FIG. 8. Transmission profiles over the phenolic layer. Receiver lines are shown with bold lines, sources by, the raypaths connecting source-receivers with thin lines and slab joints with dash lines. Receiver lines at top surface along 0, 90, 45, and 135 with respect to x 1 -axis are shown in panels (a-d). The source is located at the bottom surface. Receiver lines on the top surface along angular directions of 0, and 90 with the source also located on the top surface are shown in panels (e-f). from qp first arrival traveltimes picked on the vertical component data. The qs V and qs H group velocities are obtained from the qs-wave first arrivals picked on the radial and transverse data components. The qp and qs H first arrivals are strong and easy to pick. The qs V first arrivals, however, are more difficult to identify and pick (Figure 10b) from raw radial component data. The horizontal component of the reflected qp-wave first arrival event appears rather strong in the radial-component data, and because of its velocity (almost twice that of the shear waves) greatly interferes with the first arrivals of the qs V -wave, making the picking of the direct arrival of the qs V -wave difficult, especially for the middle-angle range. To overcome this difficulty, radial trace filtering (Henley, 2003) was applied to the radial component data. This estimate-and-subtract method attenuates the interference from events whose local dip differs from that of the qs V first arrival event. Done carefully, this type of radial filtering does not introduce traveltime changes to the target event (static shift) and also preserves the amplitude (Henley, 2003). The qp and qs velocities along the x 1 -, x 2 -, and x 3 -axes are listed in Table 2. We estimated the errors as ±70m/s and ±35m/s for the qp- and qs-velocities measured from physical model data, using 1m error in distance (scaled), 0.004s error for qp-wave first arrival time picks (1/8 of the dominant wavelength of the 32Hz wavelet), and 0.007s error for qs-waves first arrival time picks (1/8 of the dominant wavelength of the 17Hz wavelet). The qp and qs group velocity surfaces for the symmetry planes, polar plots of group velocity versus group angle are shown in Figure 9. The qs H wavefronts are purely ellipsoidal. The qp wavefronts deviate slightly from the ellipsoidal (shown by gray curves) with smaller velocities (at middle-angle range) compared to the ellipse. The qs V wavefronts also deviate from the spherical with slightly larger velocities (at middle-angles range) compared to the circle (grey curve). 10 CREWES Research Report Volume 25 (2013)

Phase and group velocity measurements Vertical group velocity (m/s) 3500 2500 1500 500 qs V qs H qp Vertical group velocity (m/s) 3500 2500 1500 500 qs V qs H qp 0 1000 2000 3000 Horizontal group velocity (m/s) (a) (x 1, x 3 ) plane 0 1000 2000 3000 4000 Horizontal group velocity (m/s) (b) (x 2, x 3 ) plane Vertical group velocity (m/s) 3500 2500 1500 500 qs V qs H qp 0 1000 2000 3000 4000 Horizontal group velocity (m/s) (c) (x 1, x 2 ) plane FIG. 9. The group velocity surfaces for the three modes (qp, qs V, qs H ) in the three symmetry planes. An elliptical wavefront is plotted for comparison in solid grey. The measured velocities in the 0 and 90 directions are considered to be the major and minor axes of the ellipse. Group angles are plotted with respect to the vertical axis for the (x 1, x 3 ) and (x 2, x 3 ) planes, and with the x 1 -axis for the (x 1, x 2 ) plane. V 11 V 22 V 33 V 23 V 13 V 12 2950 ± 70 3640 ± 70 3500 ± 70 1700 ± 35 1530 ± 35 1510 ± 35 Table 2. Phenolic qp- and qs-velocity (m/s) in principal directions. Here V ij (i, j = 1, 2, 3) is the body wave velocity which propagates along the x j -axis and is polarized along the x i -axis. For example V 11 is the qp-wave velocity propagating along the x 1 -axis, and V 23 (= V 32 ) is the qs-wave velocity propagating along the x 3 -axis and polarized along the x 2 -axis. CREWES Research Report Volume 25 (2013) 11

Mahmoudian et al. Traveltime (µs) Offset (mm) 200 100 0 100 200 20 40 60 80 100 120 direct qs V reflected qp direct qp Traveltime (µs) Offset (mm) 200 100 0 100 200 20 40 60 80 100 120 direct qp direct qsv (a) Vertical Offset (mm) 200 100 0 100 200 20 40 direct qs H (b) Radial Offset (mm) 200 100 0 100 200 20 40 direct qp direct qsv Traveltime (µs) 60 80 100 Traveltime (µs) 60 80 100 120 120 (c) Transverse (d) Radial (filtered) FIG. 10. The 3C transmission data, over the phenolic layer, acquired along the profile shown in Figure 8a, with the wave propagation at the (x 1, x 3 )-plane. (a) Raw vertical-component data. (b) Raw radial-component data. (c) Raw transverse-component data. (d) Filtered radial-component data (radial trace filtering). Red dots are first arrival picks of each mode. Displayed data have a long-gate automatic gain control applied for the vertical and transverse components. The radial component data have been displayed with a shorter window automatic gain control to boost the direct qs V arrival. The three components have similar noise levels. 12 CREWES Research Report Volume 25 (2013)

Phase and group velocity measurements FIG. 11. The group velocities versus group angle in the symmetry planes (x 1, x 3 ), (x 2, x 3 ), and (x 1, x 2 ). The solid gray lines are velocities calculated from estimated parameters and the dotted lines are measured ones. In (x 1, x 2 ) plane the black dotted line measured from a receiver profile along x 2 -axis, the magenta dotted line measured from a receiver profile along x 1 -axis. The group velocity measurements from transmission data collected by 13mm transducers, which are the commonly used transducers in physical modeling on solid surfaces, are used to estimate the experimental phenolic layer stiffness coefficients presented in Mahmoudian et al. (2013). Figure 11 compares the measured qp group velocities from transmission data collected by 1.1mm transducers and group velocities calculated from the estimated A ij ; the calculated velocities match the measured velocities reasonably well. Some small discrepancies between the theoretical group velocities and measured group velocities could be due to our assumption of homogeneity for the simulated fractured layer or the employment of the approximate orthorhombic group velocity expression rather than an exact form. The measured phase velocities are also compared to the theoretical velocities predicted from the estimated A ij (Figure 12). Good agreement within the error range of velocity measurements is obtained. Phase Velocity Measurements For each symmetry plane, we acquired two transmission seismic data gathers with the source at two different depths from the receiver plane. The P-transducers (piezoelectric pin CA-1135) with each piezoelectric element being 1.1mm in diameter were used to produce the vertical component data. Using the method discussed in Kebaili and Schmitt (1997), we measured the phase velocity for the three symmetry planes. We took great care picking on the (τ, p) transforms, a small window AGC (automatic gain control) used in order to enable consistent picking on first arrivals. The measured phase velocities are also compared to the theoretical velocities predicted from the estimated A ij (Figure 12). Good agreement within the error range of velocity measurements is obtained. This indicates that the estimated A ij from group velocity measurements can be used to obtain the phase velocities with high accuracy. CREWES Research Report Volume 25 (2013) 13

Mahmoudian et al. FIG. 12. The phase velocities versus phase angle in the symmetry planes (x 1, x 3 ), (x 2, x 3 ), and (x 1, x 2 ). The solid gray lines are theoretical velocities and the dotted lines are measured ones. In (x 1, x 2 ) plane the black dotted line was measured from a receiver profile along x 2 -axis, the light grey triangle line was measured from a receiver profile along x 1 -axis. SUMMARY AND CONCLUSIONS We obtained the group velocities in various directions through a phenolic medium with orthorhombic anisotropy. This was done by recording 3C ultrasonic transmission seismograms with 13mm diameter piezoelectric transducers as sources and receivers, picking direct arrival traveltimes, and dividing these by source-receiver nearest-edge distances. To account for large transducer diameters, we used edge-to-edge distances rather than centerto-center distances. The effective diameters of transducers were carefully determined from velocity measurements made over a known isotropic material in advance of any anisotropic experiment. Using edge-to-edge correction, the qp traveltime measurements with 13mmdiameter transducers, picked on vertical component data, were consistent with the ones made with smaller (1.1mm- and 6mm-diameter) transducers. Using very small point transducers, the measurements of group velocities obviously will not need an extra edge-to-edge correction. However, very small transducers do not usually produce a powerful enough signal for large physical models, and they have coupling issues, experimental challenges which we should be able to manage in the near future. The qs V -wave from 13mm transducers, was difficult to identify and pick on radial component data, and radial trace filtering was used to help with the picking. An alternative method commonly used for estimating A ij is based on the well-known exact relationship between phase velocity within the symmetry planes and the stiffness coefficients. The phase velocities for waves from an impulsive source can be determined by applying the (τ, p) transform to the transmission gathers with the sources at two different depths. Since the (τ, p) transform method is dependent on identifying and picking the times of unique wavefronts, the closer the transducers are to approximating point sources and point receivers, the better suited the data will be for this derivation. Large transducers appear not to be suitable for this application, because the wavefronts associated with large transducers are superpositions of energy pulses from different parts of the active faces of the transducers. Picking consistently on the strong peak or trough of the transformed wavefront 14 CREWES Research Report Volume 25 (2013)

Phase and group velocity measurements 3700 estimated first arrival peak Phase velocity (m/s) 3500 3300 3100 2900 0 30 60 Phase angle (degrees) FIG. 13. Measured qp phase velocity of phenolic layer using first arrival picks compared to picking on strong peak, from transmission data collected by 1.1mm transducers. The propagation plane is the (x 1, x 3 ) plane. will result in different trends for the phase velocities (Figure 13). Estimation of A ij from phase velocities, because of the exact formula, are potentially more accurate. However, the limitations imposed on the measurements of the phase velocities in the laboratory give the group method a definite advantage on real data. Theoretical phase and group velocities calculated using values of A ij estimated from experimental group velocities agreed with the experimental phase and group velocities, respectively, validating the A ij estimates. Our phenolic experimental layer, now with known elastic properties, is intended to be used in a modeling study for traveltime and quantitative amplitude analysis of fractured media. We assume homogeneity in our simulated fractured medium. The very slight frequency dispersion observed in the data (wavelet shapes broaden in time as source-receiver offsets increase) is likely mostly due to the wave interference caused by the large size transducers. We have demonstrated this pulse broadening numerically by finite-difference modeling of the response of an extended source shooting into an extended receiver through a non-dispersive medium (note the pulse broadening in Figure 2a). The simulation shows a stretching of the wavelet from near to far offset. Therefore, the wavelet shape change observed in physical model data should be mostly due to the dimensions of the source and receiver, and is not likely due to an intrinsic frequency dispersion effect. ACKNOWLEDGMENTS The industrial sponsors of the Consortium for Research in Elastic Wave Exploration Seismology (CREWES) and the National Science and Engineering Research Council of Canada (NSERC) provided financial support for this research. Dave Henley is acknowledged for the application of radial trace filtering on the radial component data. REFERENCES Bouzidi, Y., and Schmitt, D. R., 2009, Measurement of the speed and attenuation of the Biot slow wave using a large ultrasonic transmitter: Journal of Geophysical Research, 114, No. B08. Brown, R. J., Lawton, D. C., and Cheadle, S. P., 1991, Scaled physical modeling of CREWES Research Report Volume 25 (2013) 15

Mahmoudian et al. anisotropic wave propagation: multioffset profiles over an orthorhombic medium: Geophysical Journal International, 107, No. 3, 693 702. Dellinger, J., and Vernik, L., 1994, Do traveltimes in pulse-transmission experiments yield anisotropic group or phase velocities?: Geophysics, 59, No. 11, 1774 1779. Ebrom, D. A., and McDonald, J. A., 1994, Seismic physical modeling: Geophysics Reprint Series: Society of Exploration Geophysicists, Tulsa, OK. Henley, D. C., 2003, Coherent noise attenuation in the radial trace domain: Geophysics, 68, No. 4, 1408 1416. Kebaili, A., and Schmitt, D. R., 1997, Ultrasonic anisotropic phase velocity determination with radon transformation: Journal of the Acoustical Society of America, 101, 3278 3286. Mah, M., and Schmitt, D. R., 2001, Experimental determination of the elastic coefficients of an orthorhombic material: Geophysics, 66, No. 4, 1217 1225. Mahmoudian, F., Margrave, G. F., Daley, P. F., Wong, J., and Henley, D., 2013, Estimation of anisotropic elastic stiffness coefficients of an orthorhombic physical model using group velocity analysis on transmission data: Accepted for publication in the Geophysics. McSkimin, H. J., 1967, Physical acoustics, vol. I, Part A: Academic NewYork. Musgrave, M. J. P., 1970, Crystal Acoustics: Holden-Day, San Francisco. Tsvankin, I., 1997, Anisotropic parameters and P-wave velocity for orthorhombic media: Geophysics, 62, No. 4, 1292 1309. Vestrum, R. W., 1994, Group- and phase-velocity inversions for the general anisotropic stiffness tensor: University of Calgary, M.Sc. Thesis. Vestrum, R. W., Lawton, D. C., and Schmitt, R., 1999, Imaging structures below dipping TI media: Geophysics, 64, No. 4, 1239 1246. Wong, J., Hall, K. W., Gallant, E. V., Maier, R., Bertram, M., and Lawton, D. C., 2009, Seismic physical modeling at University of Calgary: Canadian Society of Exploration Geophysicists Recorder. 16 CREWES Research Report Volume 25 (2013)